2015 - Round 22 : Two Straight Lines

/

There were five home teams that the Head-to-Head Fund could have wagered on this week - those priced at $1.50 or more - but it opted to risk none of them, content to rest on its 2 from 3 performance over the course of the last three rounds.

That left the Line Fund responsible for the weekend's excitement and it chose to back up on the Giants, this week giving rather than receiving start, and to take a literal punt on the Saints for only the second time this season. 

Both bets being equally-sized and equally-priced means that both carry identical upside and downside, success attaching to a 1.4c gain and failure to a 1.5c loss.

(This week I'm offering a full Ready Reckoner just for old times' sake, but compressed because there's not all that much of interest to view.)

The potential impacts on the Overall Portfolio are very easy to calculate this week provided one has a proficiency with the mechanics of doubling and subtraction. Two successes add 2.8c to the value of the Portfolio, two failures strip 3c from that value, while one success and one failure leaves the Portfolio barely unchanged and just 0.1c worse off.

HEAD-TO-HEAD TIPSTERS

This week sees very similar levels of disagreement amongst the Head-to-Head Tipsters as last week, again driven up almost exclusively by the Heuristic Tipsters.

Home Sweet Home has the week's highest Disagreement Index of 50%, heading Short Term Memory I's 39%, and Shadow's 34%.

Where it counts on the MoS Leaderboard though, there is no dissension at all, the Top 6 Tipsters all selecting the nine favourites to win.

Viewed on a game-by-game basis, the Roos v Dogs game has generated the most disagreement, 13 Tipsters opting for the away and underdog Dogs, including all members of the Win, Pro and H2H families. In the Crows v Eagles game, 10 Tipsters have selected the Crows (almost all of them Heuristic Tipsters), and in the Cats v Pies game, eight Tipsters have chosen the Pies.

MARGIN PREDICTORS

The Margin Predictors, despite being unanimous about the more likely victors in six contests and almost unanimous in a seventh, have still managed to record an all-Predictor Mean Absolute Deviation (MAD) that is the highest in four weeks, and the third-highest since Round 11.

C_Marg and Combo_NN2 have the highest individual MADs once again this week though they have flipped their ordering, C_Marg recording the round high of 9.7 points per game and Combo_NN2 coming in slightly below with a 9.4 points per game result. Bookie_3 is in a relatively unaccustomed third position on this metric this week, and actually finds itself Predictor Most Extreme in two games, which is only one game fewer than C_Marg and Combo_NN2. Combo_7, for the second time in three weeks has the round's lowest MAD, its predictions no more than 7 points from the all-Predictor average in any game.

By some considerable distance, the Hawks v Lions game has the round's highest MAD, its figure of 19.4 points per Predictor a full two goals higher than the next highest. The Hawks v Lions game also sports an 11 goal range with predictions starting at an 18-point Hawthorn win (Combo_NN2) and extending all the way to an 87-point Hawthorn win (C_Marg). 

The two other games with relatively large MADs - the Cats v Pies and Roos v Dogs contests - are also the only games with more than a single Predictor on both sides of zero.

C_Marg, which currently has the lowest Mean Absolute Error (MAE) of all the Predictors this season has margin predictions no more than about two goals different from ENS_Greedy, which currently lies second, and even the worst possible set of results for C_Marg would see it surrender less than one-half of its current lead.

PROBABILITY PREDICTORS

Disagreement levels are at historically low levels this week amongst the Head-to-Head Probability Predictors, the 4.2% per game per Predictor average the fifth-lowest of the season. C_Prob, WinPred and ProPred, in that order, are the only Predictors with above-average MADs, C_Prob's heightened MAD coming on the back of being Predictor Most Extreme in six of the nine contests. WinPred and ProPred are in this same position in just three games each.

Three games have above-average MADs, the highest of all in the Roos v Dogs game where it's 9.6% points per Predictor and where the probability assessments span a 30% point range, WinPred assessing the Roos as 33% chances, and C_Prob assessing them as 63% chances.

The next-highest MAD (8.1% points per Predictor) comes in the Cats v Pies contest, and this is also the game with the second-large range of probability assessments (24% points).

The Dons v Tigers game is the one that's elicited the lowest MAD (1.6% points per Predictor), all eight of MoS' Head-to-Head Probability Predictors assessing the Dons as between 8% and 18% prospects.

Something's obviously got into the Line Fund algorithm's electronic equivalent of water this week with six teams being assessed by it as 55% or better chances in the line market, and four being assessed as 64% or better chances. The Dogs are seen as having the best prospects of all (73%), just a fraction better than the Pies (72%). I think the bonnet is going to be up on the Line Fund algorithm for some considerable time during the off-season.

FORECASTER DISAGREEMENT

Averaged across the season:

  • Home Sweet Home has the highest average weekly Disagreement Index (45.0% per round) and Easily Impressed I the next-highest (35.4% per round). ENS_Linear and ENS_Greedy have the joint-lowest average Index values (17.2 per round).
  • Combo_NN2 has the highest average MAD amongst the Margin Predictors (9.85 points per game per round) and C_Prob has the second-highest (9.57 points per game per round). Combo_7 has the lowest (2.77 points per game per round), just slightly less than Bookie_9 (2.79 points per game per round).
  • Amongst the Head-to-Head Probability Predictors, C_Prob has, by some considerable margin, the highest average MAD (8.1% points per game per round), well ahead of the second-placed WinPred (6.2% points per game per round). Bookie-OE has the lowest average MAD (4.2% points per game per round), though Bookie-RE and Bookie-LPSO both have nearly-identical averages (4.3% points per game per round).